12:28 p.m.
OK. I understand. Good to know that gretl is doing well.
It is interesting that in this url the calendar for Spain in 1700 is
different from the US calendar.
I assume the "cal" code maybe is doing some adjustments also for some
days in 1756 (and also some other years in the XVIII century). In
particular the gretl function 'weekday' produces negative values for
some days in may 1756 (they should be values only from 0(sunday) to
6(saturday)).
<script>
nulldata 1500
setobs 7 1755-01-01 --time-series
series d = weekday($obsmajor,$obsminor,$obsmicro)
</script>
El 16/02/17 a las 18:00, gretl-devel-request(a)lists.wfu.edu escribió:
Gretl follows the unix "cal" program on this. In
lib/src/calendar.c:
/* leap year -- account for gregorian reformation in 1752 */
#define leap_year(yr) \
((yr) <= 1752 ? !((yr) % 4) : \
(!((yr) % 4) && ((yr) % 100)) || !((yr) % 400))
See also
https://www.timeanddate.com/calendar/index.html?year=1700
--
Ignacio Díaz-Emparanza
Departamento de Economía Aplicada III (Econometría y Estadística)
Universidad del País Vasco - Euskalherriko Unibertsitatea, UPV/EHU
Tfno: (+34) 94 601 3732
http://www.ehu.eus/ea3
2:01 p.m.
New subject: calendar (was Re: Gretl-devel Digest, Vol 121, Issue 20)
On Thu, 16 Feb 2017, Ignacio Diaz-Emparanza wrote:
OK. I understand. Good to know that gretl is doing well.
It is interesting that in this url the calendar for Spain in 1700 is
different from the US calendar.
I assume the "cal" code maybe is doing some adjustments also for some days in
1756 (and also some other years in the XVIII century). In particular the
gretl function 'weekday' produces negative values for some days in may 1756
(they should be values only from 0(sunday) to 6(saturday)).
<script>
nulldata 1500
setobs 7 1755-01-01 --time-series
series d = weekday($obsmajor,$obsminor,$obsmicro)
</script>
Hmm, not so good. Actually the day-of-week code is taken from Uspensky
and Heaslet's Elementary Number Theory (1939) and it assumes a
Gregorian calendar, so (even with the bug you note fixed) it won't
give days of the week as they appeared on actual calendars prior to
the "cutover" from Julian to Gregorian.
I'm not sure what to do about this. One option would be to
back-project the Gregorian calendar consistently ("proleptic Gregorian
calendar"), and never mind what calendars actually said in the old
days (they would differ across countries anyway). That would, for
example, make 1700 not a leap year, though it was on the calendar in
force in England at the time. ISO 8601 has it that all dates must be
consecutive (unlike England in 1752 when Wednesday the 2nd of
September was followed by Thursday the 14th) so maybe we should make
this change.
Any other thoughts on this?
Allin
10:22 p.m.
New subject: calendar (was Re: Gretl-devel Digest, Vol 121, Issue 20)
On Thu, 16 Feb 2017, Allin Cottrell wrote:
On Thu, 16 Feb 2017, Ignacio Diaz-Emparanza wrote:
> OK. I understand. Good to know that gretl is doing well.
> It is interesting that in this url the calendar for Spain in 1700 is
> different from the US calendar.
Thanks for digging into this, Ignacio. I think we have to switch to
the proleptic Gregorian calendar. This means that our dates will not
match the historical calendar from before the introduction of the
Gregorian calendar (1752 in England; 1582 in Spain, Portugal, France
and Italy; and at various other times for other countries).
There's no way we can match all historical calendars. Up till now
we've matched the English calendar (following "cal") with regard to
leap years, but I'm afraid not with full consistency. It's extremely
difficult to achieve consistency without prolepsis: in the English
case, we'd have to deal with the fact that September 1752 had only
19 days (they skipped from Wednesday the second to Thursday the
14th)! For econometric purposes let's go with a calendar that has
consecutive dates (as I gather database software does).
At some point we could maybe add a function that translates from
Gregorian to Julian dates and vice versa, in case that's of any
interest -- I wonder how many econometricians work with archival
daily data from Julian-calendar days?
Allin
2787
days inactive
2788
days old
2 comments
2 participants
participants (2)
-
Allin Cottrell -
Ignacio Diaz-Emparanza